5D vs 20D: Wide Angle Shootout

This article and all the photographs in it is by Rom Prakash, and is licensed under the Creative Commons Attribution license.

Since I own both the Canon EF-S 10-22/3.5-4.5 and the EF 17-40/4.0L, and the EOS-20D and EOS-5D, I thought it was time for me to test them for myself. I saw some tests between the EF 16-35/2.8L and 10-22 on the Fred Miranda forum, but nothing on the 17-40 and 10-22 so far.

Testing methodology

Both cameras were mounted on tripods.
No UV filter on either lens.
Remote Cable Release used.
No mirror lock-up used (forgot to, but I don't think it has a bearing on the results). [ I agree. - Petteri ]
White Balance was set to Daylight on both Cameras.
Shot in RAW, converted in Adobe Camera RAW at default conversion settings, with no extra sharpening.

The test shots

I shot two pairs of test shots. The first is at f/11 on both lenses, and the second is at f/4. They are two separate subjects.

The f/11 shots are shot at ISO 400 as I didn't want the wind to interfere with a low shutter speed. Noise on both bodies is fairly good at ISO 400, and I'm not really interested in a noise test.

The f/4 shots were shot at ISO 100.

I tried to nudge the 10-22 a little less than half way between the 10 and 12 marking to try to get closer to 11 mm, to match the fields of view between the cameras and lenses better.

Performance stopped down

The test shots were taken at f/11, which is probably the best aperture on both lenses.

5D, full frame downsampled:

20D, full frame downsampled:

5D Bottom Left Corner:

20D Bottom Left Corner:

5D Bottom Right Corner:

20D Bottom Right Corner:

5D Center:

20D Center:

5D Upper Left Corner:

20D Upper Left Corner:

Performance at f/4

5D, full frame downsampled:

20D, full frame downsampled:

5D Close to Center:

20D Close to Center:

5D Top Left Corner:

20D Top Left Corner:

5D Detail (3/4 to the bottom of frame):

20D Detail (3/4 to the bottom of frame):

Analysis

First, pixel peeping is HARD work! I know this isn't the most scientific of rigs, and I really don't have time to do a thorough investigation. Kudos to Phil, this is mind numbing and thankless work.

What I've seen so far delights me however, and tells me that full-frame is most probably the future -- especially if Canon can tool up some better wide-angle lenses. Here are some other observations:

The 17-40 is a very usable lens on a full-frame body such as the 5D. It does behoove people to either not use a UV filter or find the slimmest filter one can fit.
I didn't notice any significant vignetting (as compared to the 20D) on the lens/body combination for this subject matter. Your mileage may vary. I'll be reshooting with my slim Heliopan on, as I have noticed more vignetting than this with it on.
Light falloff concerns with a wide-angle lens on a full-frame sensor are not a concern for me (at f/4 or below). I don't have an EF 16-35 f/2.8L lens to test with, so that's my only conclusion as far as light falloff is concerned.
The 20D/10-22 combination is a very good one.
The 5D/17-40 are outresolving the 20D/10-22 (as I'd hope!). [ Check especially the stopped-down center crops; the difference is almost dramatic. - Petteri. ]
I might be selling my 10-22 and keep my 20D for strictly telephoto/backup. The 17-40 is sufficient for my needs for WA.
The 20D LCD is brighter than the 5D's, however the 5D LCD seems to have a truer gamma/tone curve to it than the 20D. My shots are closer to my calibrated LCD on the 5D than the 20D. The RGB histogram is very nice, especially with these cameras that seem to be overly red sensitive.
I'm fairly certain the 5D has a weaker AA filter than the 20D. A lot of detail goes to mush in the 20D that the 5D can clearly pick apart. Seems like there's something more going on with the 5D sharpness than just the (fairly minor) difference in pixel count.
Both of these camera/lens combos will sharpen very nicely and provide good prints.
This isn't really a test of chromatic aberration -- but you can see in the truck with the glass on it that the CA appears to be worse on the 17-40 (reddish) than the 10-22 (greenish?)

The Galleries where you can find the originals is located at: http://rommae.smugmug.com/Pixel%20Peeping/125096.

Update: 20D crops at f/8

In response to some readers claims that f/11 is past the diffraction limit on the EOS-20D, we are providing crops from the same series shot at f/8.0. We don't see any significant improvement in them over the f/11 ones, but they're here so you can make up your own mind..

Center:

Top left:

Bottom left:

Bottom right: